The booming AI industry is consuming vast quantities of computer memory chips, the same type used in laptops, smartphones, and games consoles. This has led to higher prices and potential shortages for consumer products. Manufacturers face challenges in scaling production amid surging demand.
Artificial intelligence models require enormous amounts of memory due to their scale, consisting of billions or trillions of parameters stored in RAM to enable fast data processing and avoid bottlenecks from slower storage like hard discs. These models run at vast scales, supporting hundreds, thousands, or millions of simultaneous users, fueled by unlimited investment.
Semiconductor production is constrained by factory capacities, with new facilities requiring years and massive investments to build. Samsung Electronics and SK Hynix, which produce about 70 percent of these chips, hesitate to increase supply significantly, fearing an AI industry downturn that could leave plants idle. Samsung has capitalized on demand by raising prices, for instance, a 32-gigabyte chip from $149 in September to $239 in November, a 60 percent hike.
Similar shortages have occurred before. In 2021, a global chip crisis, exacerbated by the pandemic, trade wars, fires, droughts, and snowstorms, disrupted manufacturing from trucks to microwaves. That year also saw hard disc shortages from the rise of the Chia cryptocurrency. The AI boom has similarly hoarded graphics processing units, boosting Nvidia's share price from $13 at the start of 2021 to over $200 recently.
OpenAI has secured deals with Samsung and SK Hynix for an estimated 40 percent of global memory supply, while Microsoft, Google, and ByteDance are also acquiring chips aggressively. Equilibrium might not arrive until 2028, unless an AI bust intervenes, potentially causing economic fallout. Industry analyst Sanchit Vir Gogia of Greyhound Research warned Reuters that "the memory shortage has now graduated from a component-level concern to a macroeconomic risk."